ObjectivesRecently, comparison of drug responses on gene expression has been a major approach to identifying the functional similarity of drugs. Previous studies have mostly focused on a single feature, the expression differences of individual genes. We provide a more robust and accurate method to compare the functional similarity of drugs by diversifying the features of comparison in gene expression and considering the sample dependent variations.MethodsFor differentially expressed gene measurement, we modified the conventional t-test to normalize variations in diverse experimental conditions of individual samples. To extract significant differentially co-expressed gene modules, we searched maximal cliques among the co-expressed gene network. Finally, we calculated a combined similarity score by averaging the two scaled scores from the above two measurements.ResultsThis method shows significant performance improvement in comparison to other approaches in the test with Connectivity Map data. In the test to find the drugs based on their own expression profiles with leave-one-out cross validation, the proposed method showed an area under the curve (AUC) score of 0.99, which is much higher than scores obtained with previous methods, ranging from 0.71 to 0.93. In the drug networks, we could find well clustered drugs having the same target proteins and novel relations among drugs implying the possibility of drug repurposing.ConclusionsInclusion of the features of a co-expressed module provides more implications to infer drug action. We propose that this method be used to find collaborative cellular mechanisms associated with drug action and to simply identify drugs having similar responses.

BackgroundTyrosine kinase inhibitor (TKI)-based therapy is a recommended treatment for patients with chronic myeloid leukemia (CML). However, a considerable group of CML patients do not respond well to the TKI therapy. Challenging to overcome this problem, we tried to discover molecular signatures in gene expression profiles to discriminate the responders and non-responders of TKI therapy.MethodsWe collected three microarray datasets of CML patients having total 73 responders and 38 non-responders. Statistical analysis was performed to identify differentially expressed genes (DEGs) as gene signature candidates from integrated microarray datasets. The classification performance of these genes and further selected discriminator gene sets was tested by using random forest and iterative backward variable selection methods.ResultsWe identified a set of genes including CTBP2, NADK, AZU1, CTSH, FSTL1, and HDLBP showing the highest accuracy more than 69.44 % to classify TKI response in CML patients. Interestingly, four genes of them are on the signaling pathway of cell proliferation. This set of genes showed much higher performance than the average performance of other genes in downstream signaling of TKI target, BCR-ABL.ConclusionsIn this study, we could find a set of potential companion diagnostic markers for TKI treatment and, at the same time, the potential of gene expression analysis to enhance the coverage of companion diagnostics.Electronic supplementary materialThe online version of this article (doi:10.1186/s12920-016-0194-5) contains supplementary material, which is available to authorized users.

Large microarray data sets have recently become common. However, most available clustering methods do not easily handle large microarray data sets due to their very large computational complexity and memory requirements. Furthermore, typical clustering methods construct oversimplified clusters that ignore subtle but meaningful changes in the expression patterns present in large microarray data sets. It is necessary to develop an efficient clustering method that identifies both absolute expression differences and expression profile patterns in different expression levels for large microarray data sets. This study presents CLIC, which meets the requirements of clustering analysis particularly but not limited to large microarray data sets. CLIC is based on a novel concept in which genes are clustered in individual dimensions first and in which the ordinal labels of clusters in each dimension are then used for further full dimension-wide clustering. CLIC enables iterative sub-clustering into more homogeneous groups and the identification of common expression patterns among the genes separated in different groups due to the large difference in the expression levels. In addition, the computation of clustering is parallelized, the number of clusters is automatically detected, and the functional enrichment for each cluster and pattern is provided. CLIC is freely available at http://gexp2.kaist.ac.kr/clic.